DOI: 10.1002/cssc.201402474

Monitoring Solid Oxide CO2 Capture Sorbents in ActionChristopher J. Keturakis,[a] Fan Ni,[b] Michelle Spicer,[b] Michael G. Beaver,[b] Hugo S. Caram,[b]

and Israel E. Wachs*[a]

Introduction

The separation, capture, and storage of CO2, the major green-house gas, from industrial gas streams has received considera-ble attention in recent years due to concerns about the envi-ronmental effect of increasing CO2 concentration in the atmos-phere. The characteristics of different sorbents (zeolites, silicagels, aluminas, and activated carbons with promoters rangingfrom alkali and alkaline earth metals to amines) and their po-tential applications have recently been reviewed.[1, 2] An emerg-ing area of research utilizing reversible CO2 sorbents is thenovel sorption enhanced reaction (SER) that circumvents thethermodynamic limitations (purity and conversion) of equilibri-um-controlled reactions by selectively removing a reactionproduct from the gas phase and increasing the conversion andrate of the forward reaction.[3] For example, during the manu-facture of H2 from natural gas, the major industrial process forH2 production, CO2 sorbents are able to shift the reaction equi-librium to the right, thus,

CH4 þ 2 H2O$ 4 H2 þ CO2 ð1Þ

allowing for increased production of hydrogen in the pro-cess.[1]

Lee et al. has recently studied two promising high tempera-ture chemisorbents for the water–gas-shift-SER (WGS-SER),

a K2CO3-promoted hydrotalcite and a Na2O-promoted alumi-na.[1, 4–7] The chemisorbents were originally developed by AirProducts and Chemicals, Inc. to produce fuel-cell-grade hydro-gen by steam methane reforming.[8–11] These chemisorbentsoffer (i) reversible sorption of CO2 in the presence of steam, H2,and CO, (ii) relatively fast CO2 sorption and desorption kinetics,(iii) moderate heats of sorption (~20–65 kJ mol�1), (iv) accept-able cyclic working capacity, (v) nearly infinite selectivity of CO2

chemisorption over gases like CO, steam, N2, and CH4 at 200–500 8C, and (vi) thermal stability under WGS reaction condi-tions.

H2Oþ CO$ H2 þ CO2 ð2Þ

A comparison between the two sorbents indicated thatNa2O-promoted alumina requires about 50 % less steam for re-generation. A new analytical model for these sorbents was de-rived by Lee et al. to describe the deviation from conventionalLangmuirian equilibrium chemisorption isotherms.[4] Thismodel, though it described the data well, assumed, withoutexperimental proof, that there is one single type of adsorptionsite and that deviations from the Langmuir model are due tocomplexing between the gas and already chemisorbed CO2

molecules. These sorbents have been donated by Air Productsand Chemicals, Inc. to Lehigh University for further develop-ment.

Similar chemisorbents based on CaO (Periodic group II) andCaO-promoted alumina have also been explored by other au-thors.[1, 12–18] These CaO-based chemisorbents appear to havesimilar working capacities, tolerance to steam, and cyclic work-ing capacities, but poorer regeneration capabilities comparedto their Na2O (periodic group I)-based counterparts. At thetime of writing, only one study by Duyar et al. has demonstrat-ed that support CaO sorbents can be regenerated at tempera-tures as low as 350 8C.[18]

The separation, capture, and storage of CO2, the major green-house gas, from industrial gas streams has received considera-ble attention in recent years because of concerns about envi-ronmental effects of increasing CO2 concentration in the at-mosphere. An emerging area of research utilizes reversible CO2

sorbents to increase conversion and rate of forward reactionsfor equilibrium-controlled reactions (sorption-enhanced reac-tions). Little fundamental information, however, is knownabout the nature of the sorbent surface sites, sorbent surface–

CO2 complexes, and the CO2 adsorption/desorption mecha-nisms. The present study directly spectroscopically monitorsNa2O/Al2O3 sorbent–CO2 surface complexes during adsorption/desorption with simultaneous analysis of desorbed CO2 gas, al-lowing establishment of molecular level structure–sorption re-lationships between individual surface carbonate complexesand the CO2 working capacity of sorbents at different tempera-tures.

[a] C. J. Keturakis, Prof. I. E. WachsOperando Molecular Spectroscopy and Catalysis LaboratoryDepartment of Chemical EngineeringLehigh UniversityBethlehem, PA 18015 (USA)E-mail : iew0@lehigh.edu

[b] F. Ni, M. Spicer, Dr. M. G. Beaver, Prof. H. S. CaramDepartment of Chemical EngineeringLehigh UniversityBethlehem, PA 18015 (USA)

Supporting Information for this article is available on the WWW underhttp://dx.doi.org/10.1002/cssc.201402474.

� 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim ChemSusChem 0000, 00, 1 – 9 &1&

These are not the final page numbers! ��

CHEMSUSCHEMFULL PAPERS

Promoted aluminas for catalysis have been extensively stud-ied in the literature, though most studies focus on promotionwith transition-metal oxides rather than alkali-metaloxides.[19–27] It is now well established that transition-metaloxides are highly dispersed at low loadings on alumina andform 2D monolayers consisting of isolated sites, dimers, orpolymers by titrating the surface alumina hydroxyls. Increasingthe dopant loading above monolayer coverage (titration of allhydroxyls) ushers the formation of 3D metal-oxide nanoparti-cles. The formation of active sites for alkali and alkaline earthmetal-oxide dopants, which tend to display ionic character, isstill not well understood.

Gruene et al. characterized a CaO/Al2O3 sorbent by in situ IRspectroscopy, thermogravimetric analysis (TGA), and XRD.[15]

Neither XRD nor IR spectroscopy detected any crystallinephases, such as CaO, CaCO3, or mixed CaAl oxides, at loadingsas high as 20.3 wt % CaO/Al2O3, indicating that the calciumphase is highly dispersed with a small domain size. The TGA re-sults suggested that multiple adsorbed carbonate species existon the sorbent surface, but IR spectroscopy could not distin-guish them and only one chemisorbed carbonate speciescould be detected. Previous studies on CaO/Al2O3 also reachedthe same conclusion from Raman and IR spectroscopy, XRD,and X-ray photoelectron spectroscopy (XPS); the calciumphase is highly dispersed and not detectable as CaO, thoughminor amounts of CaCO3 are sometimes detected.[28, 29]

A recent CaO–Al2O3 mixed oxide study corroborated the sameresults.[30]

Potassium, a group I alkali metal, is expected to exhibit simi-lar surface molecular structures as sodium when impregnatedon alumina. The K2O/Al2O3 system has recently garnered re-newed attention as a NOx storage material utilized at hightemperatures and it has been shown that the potassium sur-face phases present depend on the precursor used.[31–34] A ni-trate precursor forms rhombohedral KNO3 and orthorhombicKNO3 below 300 8C. Above 300 8C, the KNO3 phase decompos-es to an amorphous phase and sometimes minor K2O is detect-ed.[33, 34] Non-nitrate precursors form K2CO3 at high calcinationtemperatures and sometimes a minor amount of K2O.[32] Forsamples calcined above 300 8C, XPS of the K 2p region re-vealed that approximately 60 % of potassium is in the K+ ionicform and approximately 40 % is in the form of a K�O bond.The molecular structure of K2O/Al2O3 calcined above 300 8C hasstill not been determined.

Early sodium-promoted alumina studies concluded thatsodium addition consumes the alumina hydroxyls, alumina hy-droxyls become more basic in character with increasingsodium loading, and Lewis acid centers are eliminated.[35–38]

Very little fundamental information is known about the Na2O/Al2O3 system such as the nature of the sorbent surface sites,sorbent surface–CO2 complexes and the CO2 adsorption/de-sorption characteristics.

In the present study we characterize Na2O/Al2O3 solid sorb-ents using in situ Raman and IR molecular vibrational spectros-copy, and TGA for determining adsorption/desorption charac-teristics and sorbent–CO2 working capacity at different temper-atures. Specific temperatures of interest are 200 and 400 8C,

which correspond to the reaction conditions of the low andhigh temperature WGS. At these temperatures physisorbedCO2 cannot exist and, thus, it is chemisorbed CO2 that is re-sponsible for the working capacity of the sorbents. The sorb-ent–CO2 surface complexes in action were monitored with op-erando temperature programmed-IR (TP-IR) spectroscopy, IRspectroscopy of the sorbent–CO2 surface complexes with si-multaneous analysis of desorbed CO2 gas with online massspectrometry (MS), allowing establishment of structure–sorp-tion relationships between individual surface carbonates andthe CO2 working capacity of the sorbents at high tempera-tures.

Results and Discussion

The 1–9 wt % Na2O/Al2O3 sorbents were synthesized by incipi-ent wetness impregnation of aqueous solutions of sodium hy-droxide (NaOH) onto g-Al2O3. Results from Brunauer–Emmett–Teller (BET) surface area analysis, given in Table 1, indicate a sys-

tematic decrease in surface area, within error, with increasingNa2O loading. The anchoring sites of Na2O on the surface hy-droxyls of the alumina were monitored with in situ IR spectros-copy and the resulting spectra of the surface hydroxyl regionare shown in Figure 1 as a function of Na2O loading. The sur-face hydroxyls on pure g-alumina have been extensively stud-ied and reviewed using modern density functional theory(DFT) calculations.[39, 40] The g-alumina employed in the presentstudy contains five distinct surface hydroxyls and their assign-ments are given in Table S1 of the Supporting Information.Upon addition of 1 % Na2O, the complete consumption of theHO-m1-AlVI surface hydroxyls (vibration at 3760 cm�1) takesplace (Note: AlV/AlVI indicates a 5/6-coordinated Al atom andm2/m3 indicates that the OH is bonded to two/three Alxx atoms).The IR intensity of the HO-m2-AlVI and HO-m3-AlVI surface hy-droxyls vibrating at 3675 and 3566 cm�1, respectively, also de-crease with Na2O addition (see Figure 1 inset). The intensity ofthe IR bands vibrating at 3734 and 3720 cm�1 remain nearlyunchanged up to 3 % Na2O, indicating a preferential consump-tion of the other surface hydroxyls. On increasing loading to5 % Na2O/Al2O3, the HO-m1-AlV and H2O-m1-AlV, the surface hy-droxyls that exhibit IR bands at 3734 and 3720 cm�1, respec-tively, are consumed. For the 7 % and 9 % Na2O/Al2O3 sorbents,the surface hydroxyls no longer change with increasing

Table 1. Sorbent calculated surface density and measured BET surfacearea.

Sample Surface density[Na atoms nm�2]

BET surface area[m2 g�1]

g-Al2O3 – 2001 % Na2O/Al2O3 0.98 1862 % Na2O/Al2O3 1.98 1633 % Na2O/Al2O3 3.00 1665 % Na2O/Al2O3 5.11 1757 % Na2O/Al2O3 7.31 1619 % Na2O/Al2O3 9.61 148

� 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim ChemSusChem 0000, 00, 1 – 9 &2&

These are not the final page numbers! ��

CHEMSUSCHEMFULL PAPERS www.chemsuschem.org

sodium oxide loading and crystalline Na2CO3 nanoparticles(NPs) appear in the corresponding in situ Raman and IR spectra(see Supporting Information, Figures S3 and S4). The experi-mental observation of Na2CO3 3D NPs above monolayer cover-age of surface carbonate species indicates saturation of thesurface Na2O monolayer. The remaining, broad surface hydrox-yl IR band at 3720 cm�1 corresponds to a thermally stablewater group that does not dissociate into a hydroxyl, H2O-m1-AlV, which is likely the reason it is not fully consumed.[40]

The vibration of the bridging Na�O�support bond is not ob-served in the Raman spectra of the Na2O/Al2O3 sorbents (seeSupporting Information, Figure S3), suggesting that the bondis ionic in character as ionic bonds are not Raman active. Givensodium’s penchant for forming ionic bonds, it seems likely thatsodium anchors by titrating the alumina surface hydroxyls bydisplacing the hydrogen and forming a Na+ ion and a counterAl�O� ion. This imparts a greater nucleophilic/basic characterto the oxygen atom. Crystalline sodium aluminate phases(main Raman bands at 440 and 490–500 cm�1) or NaOx phases(200–600 cm�1) were also not detected in the Raman spectraindicating its absence in these Na2O/Al2O3 sorbents.[41, 42]

The basic Na2O/Al2O3 sorbent was surprisingly found to al-ready contain strongly bound acidic CO2, which was capturedfrom ambient air and could not even be liberated by air calci-nation at 400 8C for 2 h. In situ Raman and IR spectroscopicanalyses show that the strongly bound surface CO2 and CO3

2�

consist of linear CO2 on all samples and two polydentate car-bonates (CO3-m3-Al), labeled polydentate-I and -II, on allsodium promoted samples, respectively (see Supporting Infor-mation, Figure S4).[43, 44] The interaction of gaseous CO2 withthe sorbents was monitored with in situ IR spectroscopy, andthe IR spectra in Figure 2 indicate that three new surface car-bonates are formed during CO2 adsorption at 200 8C. Bicarbon-ate species (HCO3–Al) are present on samples below sodiummonolayer coverage (bare Al2O3 and 1–3 % Na2O/Al2O3) withbands at 1650 (n3as), 1436 (n3s), and 1225 cm�1 (dOH).[20, 43, 45–52]

In addition, on all sodium containing samples two different bi-

dentate carbonate (OCO2-m1-Al) groups, labeled bidentate-Iand -II, are observed with bands at 1677 (n3as) + 1376 cm�1 (n3s)and 1625 (n3as) + 1311 cm�1 (n3s), respectively. It is not immedi-ately obvious that two surface bidentate species exist, howev-er, operando TP-IR data easily resolves the two surface bicar-bonates because of their different desorption kinetics, asshown below.

It is well known that CO2 (a weak acid/electrophile) adsorbsas a surface bicarbonate species via nucleophilic attack be-tween an alumina hydroxyl and CO2, which has recently beensupported by quantum chemical calculations.[53] Despite thepresence of five easily observable surface hydroxyl types(Table S1), only a single surface bicarbonate species is observedon bare g-alumina (at 200 8C) and this oxide’s CO2 working ca-pacity can increase by nearly six times with the addition ofsodium. These experimental observations indicate that not allg-alumina hydroxyls possess a strong enough basic characterat 200 8C to form surface carbonate species with CO2. Carbondioxide adsorption at room temperature on g-alumina, howev-er, has shown the presence of more than just surface bicarbon-ate species.[20, 54–57]

The operando TP-IR spectroscopy results, presented inFigure 3, provide direct monitoring of the different surface car-bonate species on the sorbent during CO2-temperature-pro-grammed desorption (TPD). In Figure 3 B, bands correspondingto a surface bicarbonate species (HCO3–Al, blue labels) de-crease dramatically between 100–200 8C, correlating very wellwith the observed 124 8C CO2 desorption peak in the massspectrum, 1 % Na2O/Al2O3 in Figure 3 A. Sorbents at monolayercoverage or above, ~5–9 % Na2O/Al2O3, possess a second CO2

desorption peak at 255 8C as well as a low temperature peak at150 8C. The 255 8C CO2 desorption peak correlates quite wellwith IR bands for the surface bidentate-II species (OCO2-m1-Al,red II-labels), Figure 3 D–F, that begin to decrease in intensityaround 200 8C and never completely disappear. Since surfacebicarbonates are no longer observed at these high Na2O load-ings, the low temperature CO2 desorption peak at 150 8C isattributed to a surface bidentate-I species (OCO2-m1-Al, redI-labels) whose bands decrease rapidly between 100–200 8C,

Figure 1. In situ DRIFTS of the hydroxyl region for dehydrated x % Na2O/Al2O3 sorbents at 200 8C. Spectra are vertically offset to aide in distinguishingsurface hydroxyls. Inset : Selected spectra stacked on top of each other todemonstrate the consumption of surface hydroxyls with increasing Na2Oloading.

Figure 2. In Situ DRIFTS of CO2 adsorption at 200 8C on x % Na2O/Al2O3 sorb-ents. The corresponding IR spectrum of each dehydrated sample was sub-tracted from all spectra during CO2 adsorption.

� 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim ChemSusChem 0000, 00, 1 – 9 &3&

These are not the final page numbers! ��

CHEMSUSCHEMFULL PAPERS www.chemsuschem.org

most easily observed in Figure 3 D and 3 F. A negative IR bandis also present at 1774 cm�1 in Figure 3 F, previously assignedto Na2CO3 NPs, indicating irrever-sible decomposition of Na2CO3

NPs that can release CO2. Onseveral sorbents it appears thatsurface bidentate-II carbonatesdo not fully desorb from the sur-face, which suggests there maybe more than just the two ob-served surface bidentate carbo-nates, with subtle differences be-tween them.

As previously discussed, theformation of surface Al�O� ionicsites in place of surface hydrox-yls, via sodium doping, impartsa greater nucleophilic/basic char-acter to the oxygen atoms ofalumina. Furthermore, we experi-mentally observe that sodiumdoping titrates all five observa-ble alumina surface hydroxylsand causes the formation of ad-ditional carbonates species withsurface bidentate (OCO2-m1-Al)and surface polydentate (CO3-m3-Al) coordinations. This is clearevidence that the addition ofsodium converts otherwise inac-tive surface hydroxyls (weak nu-cleophiles/bases) into active sur-face Al�O� , ions (strong nucleo-philes/bases) that can complexwith CO2, Scheme 1. The differ-ent Al�OH/Al�O� coordinationon g-alumina give rise to surfacecarbonates that also vary in theircoordination. In situ and operan-do spectroscopy data also revealthat some surface hydroxyls con-vert into Al�O� groups that aretoo strongly nucleophilic/basicin character, forming extremelystable surface carbonate groupsthat do not desorb at tempera-tures as high as 500 8C (surfacepolydentates).

A summary of all experimen-tally observed surface carbo-nates on g-alumina and Na2O/alumina is given in Table 2. It isdifficult to experimentally distin-guish which surface carbonatespecies bond to a specific hy-droxyl/ionic sites since severaldifferent surface hydroxyls are

consumed simultaneously upon the addition of a smallamount of sodium. It is also extremely likely that there are

Figure 3. Operando TP-IR spectroscopy of CO2 adsorption on x % Na2O/Al2O3. A) Mass spectroscopy CO2-TPDcurves from sorbents (m/e = 44), B) DRIFTS spectra of 1 % Na2O/Al2O3, C) 3 % Na2O/Al2O3, D) 5 % Na2O/Al2O3, E) 7 %Na2O/Al2O3, and F) 9 % Na2O/Al2O3. (B)–(F) have the same x-axis as (E). The dehydrated spectrum of each samplebefore CO2 adsorption was subtracted from all spectra during CO2 TPD. Bands labeled with a “*” are artifacts re-sulting from spectral subtraction.

Table 2. Summary of surface carbonates detected on the Na2O/Al2O3 sorbents.

Sorbent Species Desorptiontemperature [8C]

n3as (O�C�O)[cm�1]

n3s (O�C�O)[cm�1]

dOH

[cm�1]Dn3

[cm�1]

Al2O3 bicarbonate 124 1650 1436 1225 214

Na2O/Al2O3 polydentate-I >500 1571 1353 – 218polydentate-II >500 – 1450 – –bidentate-I 150 1672–1677 1361–1376 – 301–311bidentate-II 255 1618–1631 1311–1320 – 307–312

� 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim ChemSusChem 0000, 00, 1 – 9 &4&

These are not the final page numbers! ��

CHEMSUSCHEMFULL PAPERS www.chemsuschem.org

more surface hydroxyl types present on g-Al2O3, in addition tothe five surface hydroxyls detected in the IR spectra, especiallygiven that Digne et al. has computed over 10 unique, stablesurface hydroxyls on g-Al2O3.[39, 40]

It is hypothesized that the formation of extremely basic (nu-cleophilic) surface Al�O� ionic sites is a general phenomenonof alkali metal chemistry and metal-oxide surfaces. Early studiesexamining Li-, Na-, and K-promoted alumina found that basici-ty positively correlates with the alkali metal ionic radius, thatis, Li-alumina<Na-alumina<K-alumina.[35, 36] This hypothesis issupported by CO2 heats of adsorption of many alkali and alka-line earth metals on g-alumina, as measured by Horiuchiet al.[55] They found that all tested alkali metals (Na, K, Rb, Cs)gave the highest heats of adsorption (�160–170 kJ mol�1)compared to alkaline earth metals (Mg, Ca, Sr, Ba; �140–170 kJ mol�1) and rare earth metals (La, Ce, Pr, Nd; ~110–140 kJ mol�1) of similar loading on g-alumina (bare g-alumina~80 kJ mol�1). These data would be much more potent if themeasured surface densities of each dopant were used to nor-malize the heats of adsorption, (instead of assuming all load-ings were similar), but it makes the point that alkali metal dop-ants create surface sites with greater basicity on g-alumina.

The CO2 adsorption working capacity of the synthesizedsorbents, calculated from TGA weight analysis, is presented inFigure 4 and the maximum rate of adsorption is presented inFigure S5. The sorbents underwent three cycles of CO2 adsorp-tion/desorption at both 200 and 400 8C, to ensure accuracy ofthe measured CO2 adsorption capacities.

For an adsorption temperature of 200 8C, the CO2 adsorptioncapacity increases with Na loading and plateaus at 7 % Na2O/Al2O3 (monolayer coverage). The maximum rate of adsorptionat 200 8C parallels this trend. For adsorption at 400 8C, the CO2

adsorption capacity increases with Na loading and does notplateau. The maximum rate of adsorption at 400 8C parallelsthis trend as well. Although 200 8C is a better temperature formaximizing the overall CO2 working capacity of the sorbents,the gap between the capacity at 200 8C and 400 8C narrowssignificantly until 7 % Na2O/Al2O3, at which point a crossoverhappens. The CO2 adsorption capacity of the 9 % Na2O/Al2O3

sorbent shows significant variability at 400 8C, with a large de-crease from 0.72–0.60 mol kg�1 during the three cycles of ad-sorption/desorption.

The new molecular level insights gained from in situ and op-erando spectroscopy allows us to understand the CO2 workingcapacity trends of the Na2O/Al2O3 sorbents presented inFigure 4. Through operando TP-IR spectroscopy, it was demon-strated that the surface bicarbonate and surface bidentate-Icarbonate species are responsible for the majority of the CO2

capacity at 200 8C since bidentate-II species desorb around255 8C. An adsorption temperature of 400 8C is too high to uti-lize the surface bicarbonates and surface bidentate-I species,which would instantly desorb upon formation. Consequently,the CO2 capacity at 400 8C is generally lower. For 7 % Na2O/Al2O3, the CO2 capacities are nearly equal due to a nearly equalamount of surface bidentate-I and surface bidentate-II species,based on the desorption curve intensity of Figure 3 A. The sur-face bidentate-II species, however, desorb around 255 8C,which is much lower than the 400 8C adsorption temperature,suggesting that there are more than just these two observable,reversible surface bidentate carbonates contributing to theCO2 capacity.

At high loadings of 7 and 9 % Na2O/Al2O3, both the CO2

working capacity and maximum adsorption rate trends plateauwith a 200 8C adsorption temperature, but increase linearlywith a 400 8C temperature. The trend with a 400 8C adsorptiontemperature indicates that new, high-temperature reversiblesites are being formed on the sorbents, despite no spectral evi-dence of new surface carbonate species, only Na2CO3 3D NPs.Thus, it appears that Na2CO3 NPs provide additional, high-tem-perature reversible adsorption sites, presumably via the nega-tively charged carbonate ion (CO3

2�), however, in situ IR spec-

Scheme 1. Top: Diagram illustrating nucleophilic attack between an aluminahydroxyl and CO2 to form a bicarbonate species. Bottom: Diagram illustrat-ing nucleophilic attack between an Al�O� group and CO2 to form a biden-tate carbonate species.

Figure 4. CO2 capacity of x % Na2O/Al2O3 sorbents calculated based on TGAweight analysis at two different adsorption/desorption temperatures. Eachsorbent completed three adsorption/desorption cycles for reproducibility.

� 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim ChemSusChem 0000, 00, 1 – 9 &5&

These are not the final page numbers! ��

CHEMSUSCHEMFULL PAPERS www.chemsuschem.org

tra of CO2 adsorption on bulk Na2CO3 revealed no carbonatebands due to the bulk compounds low surface area.

Carbon dioxide adsorption on an ionic carbonate crystalshould tend to have no single adsorption site due to the equalcharge distribution around the carbon atom and the heteroge-neous nature of the surfaces of NPs, and, thus, should possessmultiple sites with different binding energies, depending onthe local environment around a negative oxygen atom. OurCO2 working capacity data support this hypothesis, which ismost evident through analysis of the shape of the CO2 TPDcurve for 9 % Na2O/Al2O3 (Figure 3 A), which is flat and nolonger has easily distinguishable desorption peaks at 155 and255 8C. Furthermore, the observed decomposition of theNa2CO3 NPs (Figure 3 F) between each adsorption–desorptioncycle (there is a 550 8C calcination step between each totalcycle) explains the decrease in capacity for 9 % Na2O/Al2O3

after three cycles. Note that the CO2 capacity at 200 8C for 7and 9 % Na2O/Al2O3 is the same because monolayer coveragehas been achieved (no new adsorption sites) and the Na2CO3

NPs possess mostly high-temperature adsorption sites. This isthe first study to monitor the dynamics of CO2 capture at hightemperatures on metal-oxide sorbents. The detailed mechanis-tic insight gained from the operando spectroscopy approachindicates that the rational synthesis of advanced metal-oxidesorbents should focus on maximizing the utilization of surfacehydroxyl groups for low temperature applications and carbon-ate nanoparticles for high temperature applications.

Conclusions

Na2O/Al2O3 sorbents were synthesized via the incipient wet-ness impregnation method and were studied by means of op-erando TP-IR spectroscopy, in situ Raman spectroscopy, andTGA analysis. Addition of the sodium promoter consumed fiveobservable g-Al2O3 surface hydroxyls and experimental evi-dence suggests the formation of (more basic) surface Al�O�

ionic sites in place of the surface hydroxyls. Sodium monolayercoverage was determined to be 6–7 wt % Na2O, with the for-mation of ionic Na2CO3 crystalline NPs seen at 7 % Na2O andabove. The in situ adsorption of CO2 revealed that bare g-alu-mina forms only reversible bicarbonate species at 200 8C be-cause most surface hydroxyls are too weakly basic to formstable carbonates. The addition of sodium and formation ofAl�O� ionic sites allowed many more hydroxyl sites to be ac-cessed for CO2 capture, with Na2O/Al2O3 sorbents forming bi-dentate and polydentate carbonates. It is hypothesized thathydroxyl conversion into basic surface M-O� ionic sites is a gen-eral phenomenon arising from the ionic nature of alkali metalchemistry.

Operando TP-IR spectroscopy revealed that bicarbonates, bi-dentate-I, and bidentate-II carbonate species are responsiblefor CO2 desorption peaks at 124, 150, and 255 8C respectively,and the majority of the CO2 working capacity of the sorbentswhen utilized at 200 8C. Experimental data suggests, however,that there could be more than just the three major reversiblecarbonates that were observed. Additionally, it was observedthat surface ionic Na2CO3 NPs possess multiple adsorption sites

with different binding energies, due to their heterogeneousnature, which can reversibly adsorb CO2 at higher tempera-tures than Al�O� sites. These NPs can also thermally decom-pose at high temperatures, irreversibly releasing CO2. Thesenew insights into the CO2 capture mechanism indicate that therational synthesis of advanced metal-oxide sorbents shouldfocus on maximizing the utilization of surface hydroxyl groupsfor low-temperature applications and carbonate nanoparticlesfor high-temperature applications.

Experimental Section

Sorbent synthesis and preparation : The supported Na2O aluminasorbents were prepared by incipient wetness impregnation ofaqueous solutions of sodium hydroxide (NaOH, Fisher Scientific,97.8 %) and deionized water on an alumina support (g-Al2O3,200 m2 g�1, Engelhard). The alumina support underwent calcinationat 500 8C for 4 h in air (Airgas, Zero grade) prior to impregnation.Samples were impregnated with sodium loadings of 1, 2, 3, 5, 7,and 9 wt % Na2O. The samples were allowed to dry overnightunder ambient conditions, followed by a second drying step inflowing air (100 mL min�1) at 100 8C for 4 h in a programmable fur-nace (Thermolyne, Model 48000). Finally, the samples were subject-ed to calcination by ramping the temperature at 5 8C min�1 underflowing air (Airgas, Zero grade) to 400 8C for 2 h. The final synthe-sized sorbents are denoted as x % Na2O/Al2O3, where x = weight %of sodium oxide.Surface area analysis: The quantity of sample used for the adsorp-tion experiments was chosen to match an approximate samplesize of 1000 m2 g�1. Measurements were performed using the Mi-cromeritics ASAP 2020 Surface Area and Porosity Analyzer. The cat-alysts were initially degassed for 12 h at 350 8C before adsorptionmeasurements. For surface area analysis, a N2-adsorption studywas performed. An adsorption isotherm was obtained at differentrelative pressures (P0/P from 0.01 to 1.0) with the sample sub-merged in liquid N2. The BET method was used to obtain the spe-cific surface area for the samples and was conducted using thesoftware provided by Micromeritics.CO2 sorption working capacity and maximum adsorption rate :A TA instruments SDT Q600 was used for the TGA analysis. Nitro-gen (Praxair, >99.998 %) was used as the purge gas and CO2 (Prax-air, >99.998 %) for adsorption steps. A typical TGA run was carriedout with around 30 mg powder using platinum pans at 1 atm. Ina typical cycle, the sample was first heated at 5 8C min�1 to 550 8Cand kept at that temperature for 4 h in 100 mL min�1 flowing N2 inorder to dehydrate the sample and desorb residual CO2. Duringthe adsorption step the temperature was equilibrated to 400 8Cand the gas was switched to 60 mL min�1 flowing CO2 for 4 h. Theinfluence of flow rate change is negligible and was confirmed bychecking the weight signal change using empty platinum pans.During the desorption step the gas was switched to 100 mL min�1

flowing N2 at the same temperature for 5 h. With a N2 purge, thetemperature was equilibrated at 200 8C and the gas was switchedto 60 mL min�1 flowing CO2 for 4 h, followed by desorption underN2 for 5 h. These steps outline one adsorption/desorption cycle.Each of the samples was tested for the three complete cycles.The maximum adsorption rate of CO2 was determined from the de-rivative of the TGA weight versus time curve for only the firstcycle. The derivative was calculated using a seven point Savitsky–Golay method implemented in Microsoft Excel. The maximum de-rivative value was taken as the maximum adsorption rate.

� 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim ChemSusChem 0000, 00, 1 – 9 &6&

These are not the final page numbers! ��

CHEMSUSCHEMFULL PAPERS www.chemsuschem.org

In situ diffuse reflectance infrared Fourier transform spectrosco-py (DRIFTS) of dehydrated sorbents: The in situ IR spectra of thesorbents were obtained with a Thermo Nicolet 8700 FT-IR spec-trometer equipped with a Harrick Praying Mantis attachment(model DRA-2) for diffuse reflectance spectroscopy. Approximately30 mg of each sample (powder) was loaded into the cup of thehigh temperature Harrick cell (model HVC-DR2 with a ZnSewindow). The sample was initially heated at a rate of 10 8C min�1

(Harrick ATC/low voltage temperature control unit) in the Harrickcell to 400 8C and held for 45–60 min under 30 mL min�1 flowing10 % O2/Ar (Airgas, certified, 10.00 % O2/Ar balance) before coolingto 200 8C. The in situ IR spectra were collected at 200 8C using anMCT detector (cooled with liquid N2) with a resolution of 4 cm�1

and an accumulation of 72 scans (�90 s per scan).The collection of the initial IR gas phase background was per-formed by placing a reflective mirror in the laser path, while usingthe Harrick Praying Mantis attachment, at approximately the sameheight as a full sample cup of the Harrick Cell. A new backgroundwas taken for each sample.In situ DRIFTS of CO2 adsorption: The adsorption of CO2 was per-formed after collecting a DRIFTS spectrum of each dehydratedsample at 200 8C. This adsorption temperature was chosen basedon sorbent operating temperatures in SER technology.[1] After thein situ DRIFTS spectrum was collected at 200 8C, the gas wasswitched to 5 mL min�1 of flowing CO2 (Airgas, bone dry grade).The sample was held in this environment for one hour. A macrowas made using the program OMNIC Macros/Basic to continuouslycollect spectra for the full hour. With a resolution of 4 cm�1 and72 scans, each spectrum took approximately one minute and thirtyseconds to collect, giving roughly 40 spectra for the full hour. TheDRIFTS spectrum of each dehydrated sample was subtracted fromall 40 spectra (performed in OMNIC, subtraction factor = 1) so thatonly the vibrations of the adsorbed COx species are observed.Operando temperature programmed-IR spectroscopy: A massspectrometer (Pfeiffer Quadrupole Mass Spectrometer, ModelQME200) was attached to the outlet stream of the Harrick cell inorder to monitor the exiting gas phase while simultaneously ob-taining in situ DRIFTS spectra of the sample. The mass specvacuum was maintained around 5.1 � 10�9 atm. The activation andCO2 adsorption procedure used were identical to that previouslyreported above. After a full hour of CO2 adsorption at 200 8C, thesample was allowed to cool to 50 8C under the flowing CO2 atmos-phere. Once at 50 8C, the gas was switched to argon (Airgas, UHP)at a flow of 30 mL min�1. The sample was maintained at 50 8Cunder the argon flow for ~30 min to help eliminate any physisor-bed species and allow the mass spec to reach a baseline value foreach mass monitored. Afterwards, the sample temperature wasramped from 50 8C to 500 8C at 10 8C min�1. The mass spec was setto continuously analyze the exiting CO2 gas and the IR system wasset to record a spectrum every ~90 s. The corresponding in situDRIFTS spectra of the dehydrated sorbents were subtracted fromall spectra taken during CO2-TPD (performed in OMNIC, subtractionfactor = 1). This allows for spectral examination of the desorptionbehavior of only the adsorbed species.

Acknowledgements

We would like to thank Professor Mark A. Snyder of Lehigh Uni-versity for use of TGA equipment and Air Products Inc. for the do-nated industrial sorbents for comparison. Funding was providedby the Department of Energy (DOE), grant DE-PS26-O4NT-42454,

the Pennsylvania Infrastructure Technology Alliance (PITA), grantsPITA-442-04 & PITA-542-5, and NSF-REU grant #0609018.

Keywords: carbon dioxide capture · IR spectroscopy ·operando spectroscopy · sorbent · surface chemistry

[1] K. Lee, M. Beaver, H. Caram, S. Sircar, Ind. Eng. Chem. Res. 2008, 47,8048.

[2] S. Choi, J. H. Drese, C. W. Jones, ChemSusChem 2009, 2, 796 – 854.[3] B. T. Carvill, J. R. Hufton, M. Anand, S. Sircar, AIChE J. 1996, 42, 2765.[4] K. B. Lee, M. G. Beaver, H. S. Caram, S. Sircar, AIChE J. 2007, 53, 2824 –

2831.[5] K. Lee, M. Beaver, H. Caram, S. Sircar, Adsorption 2007, 13, 385.[6] K. B. Lee, M. G. Beaver, H. S. Caram, S. Sircar, J. Power Sources 2008, 176,

312 – 319.[7] M. G. Beaver, H. S. Caram, S. Sircar, J. Power Sources 2010, 195, 1998 –

2002.[8] T. R. Gaffney, T. C. Golden, S. G. Mayorga, J. R. Brzozowski, F. W. Taylor,

1999.[9] J. R. Hufton, S. G. Mayorga, S. Sircar, AIChE J. 1999, 45, 248 – 256.

[10] S. Sircar, C. Golden, 2001.[11] W. E. Waldron, J. R. Hufton, S. Sircar, AIChE J. 2001, 47, 1477.[12] S. K. Bhatia, D. D. Perlmutter, AIChE J. 1983, 29, 79 – 86.[13] B. Balasubramanian, A. L. Ortiz, S. Kaytakoglu, D. P. Harrison, Chem. Eng.

Sci. 1999, 54, 3543 – 3552.[14] A. A. G. Belova, T. M. Yegulalp, C. T. Yee, Energy Procedia 2009, 1, 749 –

755.[15] P. Gruene, A. G. Belova, T. M. Yegulalp, R. J. Farrauto, M. J. Castaldi, Ind.

Eng. Chem. Res. 2011, 50, 4042 – 4049.[16] Z. Zhou, Y. Qi, M. Xie, Z. Cheng, W. Yuan, Chem. Eng. Sci. 2012, 74, 172 –

180.[17] N. Amos, M. Widyawati, S. Kureti, D. Trimis, A. I. Minett, A. T. Harris, T. L.

Church, J. Mater. Chem. A 2014, 2, 4332 – 4339.[18] M. S. Duyar, R. J. Farrauto, M. J. Castaldi, T. M. Yegulalp, Ind. Eng. Chem.

Res. 2014, 53, 1064 – 1072.[19] M. A. Vuurman, I. E. Wachs, J. Phys. Chem. 1992, 96, 5008.[20] A. M. Turek, I. E. Wachs, E. Decanio, J. Phys. Chem. 1992, 96, 5000.[21] I. E. Wachs, Colloids Surf. A 1995, 105, 143.[22] I. E. Wachs, Catal. Today 1996, 27, 437 – 455.[23] X. Liu, R. Truitt, J. Am. Chem. Soc. 1997, 119, 9856.[24] I. E. Wachs, Y. Chen, J.-M. Jehng, L. E. Briand, T. Tanaka, Catal. Today

2003, 78, 13 – 24.[25] I. E. Wachs, Catal. Today 2005, 100, 79 – 94.[26] G. Cr�peau, V. Montouillout, A. Vimon, L. Mariey, T. Cseri, F. Mauge, J.

Phys. Chem. B 2006, 110, 15172.[27] I. E. Wachs, C. J. Keturakis in Monolayer Systems, Vol. 7 (Ed. : R. Schlçgl),

Elsevier, 2013.[28] A. R. Saini, B. G. Johnson, F. E. Massoth, Appl. Catal. 1988, 40, 157 – 172.[29] V. Bolis, G. Magnacca, C. Morterra, Res. Chem. Intermed. 1999, 25, 25 –

56.[30] H. Dathe, A. Jentys, P. Haider, E. Schreier, R. Fricke, J. A. Lercher, Phys.

Chem. Chem. Phys. 2006, 8, 1601 – 1613.[31] X. Wang, I. E. Wachs, Catal. Today 2004, 96, 211 – 222.[32] F. Prinetto, M. Manzoli, S. Morandi, F. Frola, G. Ghiotti, L. Castoldi, L.

Lietti, P. Forzatti, J. Phys. Chem. C 2010, 114, 1127 – 1138.[33] D. H. Kim, K. Mudiyanselage, J. Szanyi, J. C. Hanson, C. H. F. Peden, J.

Phys. Chem. C 2014, 118, 4189 – 4197.[34] M. E. G�lvez, S. Ascaso, P. Stelmachowski, P. Legutko, A. Kotarba, R. Mo-

liner, M. J. Lazaro, Appl. Catal. B 2014, 152 – 153, 88 – 98.[35] R. Fiedorow, I. G. Dalla Lana, J. Phys. Chem. 1980, 84, 2779 – 2782.[36] P. O. Scokart, A. Amin, C. Defosse, P. G. Rouxhet, J. Phys. Chem. 1981, 85,

1406 – 1412.[37] V. B. Kazansky, A. V. Zaitsev, V. Y. Borovkov, A. L. Lapidus, Appl. Catal.

1988, 40, 17 – 25.[38] V. A. Ivanov, A. Pieplu, J. C. Lavalley, P. Nortier, Appl. Catal. A 1995, 131,

323 – 334.[39] M. Digne, P. Sautet, P. Raybaud, P. Euzen, H. Toulhoat, J. Catal. 2002,

211, 1 – 5.

� 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim ChemSusChem 0000, 00, 1 – 9 &7&

These are not the final page numbers! ��

CHEMSUSCHEMFULL PAPERS www.chemsuschem.org

[40] M. Digne, P. Sautet, P. Raybaud, P. Euzen, H. Toulhoat, J. Catal. 2004,226, 54 – 68.

[41] H. R. Watling, P. M. Sipos, L. Byrne, G. T. Hefter, P. M. May, Appl. Spectrosc.1999, 53, 415 – 422.

[42] D. McKeown, Am. Mineral. 2005, 90, 1506 – 1517.[43] G. Busca, V. Lorenzelli, Mater. Chem. 1982, 7, 89.[44] P. Nortier, P. Fourre, A. B. M. Saad, O. Saur, J. C. Lavalley, Appl. Catal.

1990, 61, 141 – 160.[45] J. B. Peri, J. Phys. Chem. 1966, 70, 3168 – 3179.[46] N. D. Parkyns, J. Chem. Soc. A 1969, 410 – 417.[47] E. Baumgarten, A. Zachos, Spectrochim. Acta 1981, 37A, 93 – 98.[48] D. G. Rethwisch, J. A. Dumesic, Langmuir 1986, 2, 73 – 79.[49] D. G. Rethwisch, J. A. Dumesic in Adsorptive and Catalytic Properties of

Carbon Monoxide and Carbon Dioxide Over Supported Metal Oxides,Vol. 363 (Ed. : W. M. Ayers), American Chemical Society, 1988, pp. 102 –122.

[50] M. Cabrejas Manchado, J. M. Guil, A. P. Masia, A. R. Paniego, J. M. T.Menayo, Langmuir 1994, 10, 685 – 691.

[51] H. Zou, X. Ge, J. Shen, Thermochim. Acta 2003, 397, 81 – 86.[52] A. Vimont, J. C. Lavalley, A. Sahibed-Dine, C. O. Arean, M. R. Delgado, M.

Daturi, J. Phys. Chem. B 2005, 109, 9656 – 9664.[53] J. Baltrusaitis, J. H. Jensen, V. H. Grassian, J. Phys. Chem. B 2006, 110,

12005 – 12016.[54] P. Galhotra, 2010.[55] T. Horiuchi, H. Hidaka, T. Fukui, Y. Kubo, M. Horio, K. Suzuki, T. Mori,

Appl. Catal. A 1998, 167, 195 – 202.[56] J. I. Di Cosimo, V. K. Diez, M. Xu, E. Iglesia, C. R. Apesteguia, J. Catal.

1998, 178, 499 – 510.[57] S. Walspurger, L. Boels, P. Cobden, G. Elzinga, W. Haije, R. van den Brink,

ChemSusChem 2008, 1, 643.

Received: May 31, 2014

Revised: August 22, 2014

Published online on && &&, 0000

� 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim ChemSusChem 0000, 00, 1 – 9 &8&

These are not the final page numbers! ��

CHEMSUSCHEMFULL PAPERS www.chemsuschem.org

FULL PAPERS

C. J. Keturakis, F. Ni, M. Spicer,M. G. Beaver, H. S. Caram, I. E. Wachs*

&& –&&

Monitoring Solid Oxide CO2 CaptureSorbents in Action

Tailorable, reversible solid oxide CO2

sorbents : Sodium-doped Al2O3 solidoxide sorbents are synthesized andmonitored in action, during CO2 cap-ture, with operando IR spectroscopy.A structure–sorption relationship is es-tablished between sorbent–CO2 surfacecomplexes and the sorbent working ca-pacity. Surface Al2O3 hydroxyl groupsare responsible for low temperature re-versible adsorption while small carbon-ate nanoparticles are responsible forhigh temperature adsorption.

� 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim ChemSusChem 0000, 00, 1 – 9 &9&

These are not the final page numbers! ��